Alan Hargreaves' Weblog

The ramblings of an Australian SaND TSC* Principal Field Technologist

* Solaris and Network Domain Technology Support Centre - The group I work for

Tags

(update 1) acoustic bind birthday blues bugs cec cec2007 cec2008 china cmt contention cringley debugging dogs dtrace earthquake encumbered-binaries extra flash funny google guitar halloween huron install kids linux liveupgrade locking mdb music mysql newyear niagra openjava opensolaris oracle patches patents percussion performance redhat secondlife security solaris sru sun support sxcr t2 t2000 timeslider ufs upgrade virtualbox windows youtube zfs
« Brief report on... | Main | Sun Microsystems... »
pageicon Saturday Aug 27, 2005

Comment on OpenSolaris bugfix process in a comment in Jim's blog

There was a comment attached to one of Jim's blog articles from sdestika referring to what he perceives as unnecessary bureaucracy to fix bugs in OpenSolaris.

As the author of the article and the presentation that they refer to, I have a few comments.

  1. Some of the steps currently involved are more complex than they need to be simply because we do not have an externally facing SCM. This is a bug, it's being worked on.
  2. Solaris has a reputation for quality and stability. Part of the "bureaucracy" that you refer to are the steps that we already take internally to guarantee that we don't compromise that quality. Just because we have open sourced the product does not mean that we should than sacrifice the checks and balances needed to maintain that quality.
  3. Which would you have us do?
    1. Simply say that as we don't have a method for developers outside of Sun to contribute, so tough cheese until we do?
    2. Add a little extra process to take this into account to allow those that want to contribute to do so?

Also, I make a lot of comments about these slides in the mp3 that I also left there, have you had a listen to that?

Yes, our processes are not perfect, but the important thing is, it's possible to work with. We are working to simplify things, but please don't expect us to sacrifice quality for expedience.

Alan.

Technorati Tag:

Update

I've started a thread at the Open Solaris General discussion forum. The actual thread is Bug fix process overly burdonsome ?. It would be nice to move discussion there for a wider audience. Perhaps we do need anotehr bugfix process for those who don't want to be involved in the putback processes. It would be interesting to see some good discussion about what parts of the current process are overly burdonsome.

Posted at 10:53AM Aug 27, 2005 by Alan Hargreaves in OpenSolaris  |  Comments[3]

Comments:

When did I say kill quality? None of the "methodical" processes you mentioned are _necessary_ for quality. Do you think other OSS projects who maintain quality have these kind of silly bureaucratic processes?

Convenience is the word - On LKML, I can send an email in certain format and include a patch - that's it. It gets accepted and tested eventually only if it meets certain standards. The whole process is so friendly and allows you as a developer to focus on crux of the matter. A genius techie (not an employee) will not bother with what you guys are expecting him/her to do - only a paid for employee will. Remember you are trying to garner support of OSS developers whom you aren't paying. You have to mix with their culture rather than trying to enforce things on them. That's just not going to work.

People will find their way out of Sun's control and bureaucracy by forking the code which is not a good thing IMHO.

I guess it is intended to go that route - so it's per plan for Sun - just not the OSS culture, speed and volume. (Java - Sun claims it to be open, but to date we don't have a Java UI in Sun distributed JDK which doesn't suck (Universal truth - no argument accepted) - There are thousands asking for it but Sun just isn't able to deliver on that front. If it was truly a community driven project on the scale of Linux Kernel, KDE, GNOME - that problem would have been solved a long before.)

Even in your reply the focus is on "we, Sun" rather than community. Instead of the community figuring out what to do for a "external facing SCM" you guys will go thru thousand procedural steps and eventually put up something which isn't so appealing to the community or doesn't suit their needs. Now look at how LKML solved it - within a month of bitkeeper mishap 3 different SCMs were coded and people started using what they liked - Git/Mercurial etc. And it all works well together. Everyone is happy.

Quality can't come in between hacker spirit and the competence to deliver things - it's an end result of passion and enthusiasm. You kill either and it becomes a routine day to day job rather than the mind boggling, unbounded spirit of OSS.

Sun just doesn't get it - said a thousand times and more true every time.

Posted by sdestika on August 29, 2005 at 02:24 PM EST #

OK, I had to make an assumption about what parts of bureaucracy you were commenting about because your initial comment was not clear.

The point that I am making is that the processes already exist for a reason, that is that we do not have a small number of people who put back into the code tree, the checks and balances are to ensure that we maintain the quality.

Those of us who have previously worked in the ON code tree understand why these checks and balances are present and are happy to work within them.

My understanding (which could be completely wrong and I'm sure that you'll correct me if I am) is that the people who do the actual code tree modifications to the Linux kernel are a relatively small group who have a great understanding of how everything hangs together. I imagine that a lot of this stuff goes on behind the scenes, perhaps less formally as can be done with a smaller group (perhaps not less formally - I have not seen the process I can't really comment).

Part of open sourcing of Solaris is that we want developers to be able to do the code tree modifications, just as Sun has mny thousands of engineers who do the code tree modifications.

The process that I outlined is what is required to get to the actual putback. In the Linux world I suspect (again see my prior note about being corrected), that such processes are hidden from the person who sent out the mail to lkml.

I don't work with Java, it's not the topic of the discussion, I'm not going to talk about it.

On the subject of the SCM, where did I say that "we" referred to Sun? If you check out the opensolaris-discuss list, you will see that the community has been having a huge discussion on this very issue.

Once the bug tool has been fixed to allow for updates, there will be nothing stopping anyone simply adding a comment to a bug with a suggsted fix, and having someone else do the background work.

OK, now that we can see that we have both made different assumtions about what each other was saying, how about you explain to me exactly what parts of the process you find overly burdonsome?

Posted by Alan Hargreaves on August 29, 2005 at 02:41 PM EST #

Sdestika, I've updated this blog article with a link to the open solaris discussion forum on this topic. I'd really appreciate if you could take part in this and who knows, maybe some good will come of a rational discussion.

Alan.

Posted by Alan Hargreaves on August 29, 2005 at 03:52 PM EST #

Post a Comment:
Comments are closed for this entry.